Do the different ways of investing in gold produce similar outcomes? In their September 2011 paper entitled “A Comparative Analysis of the Investment Characteristics of Alternative Gold Assets”, Tim Pullen, Karen Benson and Robert Faff examine the diversification, hedging and safe haven properties of gold bullion, ten gold stocks, 11 gold mutual funds and two gold exchange traded funds (ETFs). A diversifier exhibits a positive (but less than one) average correlation with a reference asset/portfolio. A strong (weak) hedge exhibits negative (zero) average correlation with a reference asset/portfolio. A strong (weak) safe haven exhibits negative (zero) correlation with a reference asset/portfolio during market crises. They consider non-linearity by amplifying or pre-selecting intervals of extreme negative returns for the reference asset. Using daily levels of alternative gold assets and the S&P 500 Total Return Index as a reference asset during July 1987 through June 2010 (for bullion and gold mutual funds) and February 2005 through June 2010 (for all gold alternatives), they find that:
- Gold bullion is generally a strong hedge for equities rather than a diversifier.
- Gold stocks, gold mutual funds and gold ETFs tend to be diversifiers for equities rather than hedges (with the Central Fund of Canada a notable exception as a strong hedge and safe haven).
- Gold bullion and gold ETFs act as safe havens for equities.
In summary, investors interested in gold as a safe haven from (diversifier for) equities should utilize bullion or gold ETFs (gold stocks or gold mutual funds).
Cautions regarding findings include:
- While the sample period includes several bull and bear markets for equities and several crises, it encompasses essentially a single long-term disinflationary environment. A long-term acceleration in inflation might affect correlations.
- The study identifies market crises either retrospectively or via thresholds based on the full sample (rather than inception-to-date data). An investor operating in real time may have identified crises somewhat differently.
- The statistical significance test assumes tame return distributions. To the extent the distributions are wild, this test may mislead.
